Helping vulnerable, diverse students detect and combat fake news via artificial intelligence dashboards Seminar

Fake news can kill. Some believed Covid-19 fake news, rejected vaccines, got infected, and died. Others believed stolen USA presidential election tweets, stormed the Congress barricades, and died.

Integrating four theories (formal linguistics, politeness, information market, situational theory of problem solving) yields deceptive writing strategies theory. We test this theory on 4,165 Covid-19 tweets and 31,128 “stolen election” tweets by combining several methods (machine learning, computational linguistics, multilevel diffusion analysis).

Tweets with first person singular pronouns or third person pronouns were often true, but tweets that were rude or with second person pronouns were often false. Effects of common words, emotions, and hedges differed across topics. Unlike true tweets, false tweets started earlier, spread faster, and spread further via greater broadcast influence (e.g., Trump tweet) and greater word-of-mouth influence.

Synthesizing these theories, methods, and results into our artificial intelligence dashboards helps students, organizations and governments detect and combat fake news.